AW: [ADSM-L] AW: [ADSM-L] disk layout for db and log
Hi, at the Oxford TSM Symposium there was a contribution by Jarrett Potts about improving performance: http://tsm-symposium.oucs.ox.ac.uk/papers/Finding%20Performance%20Bottlenecks%20in%20Tivoli%20Storage%20Manager%20Environments%20(Jarrett%20Potts).pdf We use raw logical db (16 GB) and log volumes (4 GB) and mirror with TSM. best regards, Kirsten Glöer Forschungszentrum Karlsruhe GmbH Institut für Wissenschaftliches Rechnen (IWR) Hermann-von-Helmholtz-Platz 1 Tel.: +49 (0)7247-82-8616 76344 Eggenstein-LeopoldshafenFax: +49 (0)7247-82-4972 Deutschland e-mail: [EMAIL PROTECTED] -Ursprüngliche Nachricht- Von: ADSM: Dist Stor Manager [mailto:[EMAIL PROTECTED] Im Auftrag von Dirk Kastens Gesendet: Mittwoch, 19. Oktober 2005 17:32 An: ADSM-L@VM.MARIST.EDU Betreff: Re: [ADSM-L] AW: [ADSM-L] disk layout for db and log Hi, Gloeer, Kirsten schrieb: > within the performance guide it is recommended to use multiple db > volumes (4 - 16) and multiple log volumes. Where did you get the information about using 4-16 volumes? The performance guide only talks about "multiple" volumes on "multiple" disks. > A TSM server database with 195 GB seems to be very large in order to > reach high performance rates. At the moment our database is 48 GB in size and is 65% utilized. What, if I use raw logical volumes for the db and log volumes? Maybe I should start with 3 or 4 rlv of size 20 GB, each on a separate disk and let them be mirrored by TSM. My other question was about allocationg db and log volumes on the same disks or separating them. So there are two possible configurations: hdisk2: db1 (20GB) + log1 (4GB) -> mirror on hdisk5 hdisk3: db2 (20GB) + log2 (4GB) -> mirror on hdisk6 hdisk4: db3 (20GB) + log3 (4GB) -> mirror on hdisk7 or hdisk2: db1 (20GB) + db2 (20GB) -> mirror on hdisk5 hdisk3: db3 (20GB) + db4 (20GB) -> mirror on hdisk6 hdisk4: log1 (4GB) + log2 (4GB) + log3 (4GB) -> mirror on hdisk7 -- Regards, Dirk Kastens Universitaet Osnabrueck, Rechenzentrum (Computer Center) Albrechtstr. 28, 49069 Osnabrueck, Germany Tel.: +49-541-969-2347, FAX: -2470
Re: AW: [ADSM-L] disk layout for db and log
Hi, Gloeer, Kirsten schrieb: within the performance guide it is recommended to use multiple db volumes (4 - 16) and multiple log volumes. Where did you get the information about using 4-16 volumes? The performance guide only talks about "multiple" volumes on "multiple" disks. A TSM server database with 195 GB seems to be very large in order to reach high performance rates. At the moment our database is 48 GB in size and is 65% utilized. What, if I use raw logical volumes for the db and log volumes? Maybe I should start with 3 or 4 rlv of size 20 GB, each on a separate disk and let them be mirrored by TSM. My other question was about allocationg db and log volumes on the same disks or separating them. So there are two possible configurations: hdisk2: db1 (20GB) + log1 (4GB) -> mirror on hdisk5 hdisk3: db2 (20GB) + log2 (4GB) -> mirror on hdisk6 hdisk4: db3 (20GB) + log3 (4GB) -> mirror on hdisk7 or hdisk2: db1 (20GB) + db2 (20GB) -> mirror on hdisk5 hdisk3: db3 (20GB) + db4 (20GB) -> mirror on hdisk6 hdisk4: log1 (4GB) + log2 (4GB) + log3 (4GB) -> mirror on hdisk7 -- Regards, Dirk Kastens Universitaet Osnabrueck, Rechenzentrum (Computer Center) Albrechtstr. 28, 49069 Osnabrueck, Germany Tel.: +49-541-969-2347, FAX: -2470
AW: [ADSM-L] disk layout for db and log
Hi, within the performance guide it is recommended to use multiple db volumes (4 - 16) and multiple log volumes. A TSM server database with 195 GB seems to be very large in order to reach high performance rates. Isn't it better to split the TSM server in two servers? best regards, Kirsten Glöer Forschungszentrum Karlsruhe GmbH Institut für Wissenschaftliches Rechnen (IWR) Hermann-von-Helmholtz-Platz 1 Tel.: +49 (0)7247-82-8616 76344 Eggenstein-LeopoldshafenFax: +49 (0)7247-82-4972 Deutschland e-mail: [EMAIL PROTECTED] -Ursprüngliche Nachricht- Von: ADSM: Dist Stor Manager [mailto:[EMAIL PROTECTED] Im Auftrag von Dirk Kastens Gesendet: Mittwoch, 19. Oktober 2005 13:52 An: ADSM-L@VM.MARIST.EDU Betreff: [ADSM-L] disk layout for db and log Hi, we bought a new IBM @server p5 520 for TSM. The server has 8 internal 70 GB harddisks. hdisk0 and hdisk1 contain the mirrorred rootvg with the operating system. So we have six disks left for the TSM db and recovery log. Both shall be mirrored, what makes three disks. What is the best disk layout? Should we create one big db volume on each of the three disks plus one big log volume? Maybe hdisk2: db1 (65 GB) + log1 (4 GB) hdisk3: db2 (65 GB) + log2 (4 GB) hdisk4: db3 (65 GB) + log3 (4 GB) Or should the recovery log be located on a separate disk (what will be a waste of space)? Shall we create on big db volume on every disk or several small volumes? Maybe hdisk2: db1 + db2 + db3 (20 GB) + log1 (10 GB) hdisk3: db4 + db5 + db6 (20 GB) hdisk4: db7 + db8 + db9 (20 GB) Any ideas are welcome. -- Regards, Dirk Kastens Universitaet Osnabrueck, Rechenzentrum (Computer Center) Albrechtstr. 28, 49069 Osnabrueck, Germany Tel.: +49-541-969-2347, FAX: -2470
Re: disk layout for db and log
Jim, Are you going to mirror at the hardware, OS, or application level? I'm going to mirror at the application (TSM) level. Spread the logical volumes across multiple disks at least two. The database, logs, and storage pool volume need to run on their own disks. The stgpool volumes are located in our SAN. If the database and log volumes go on separate disks, I would only have two disks for the db and one for the small recovery log (which are mirrored to the other three disks). I think it could be possible to have on disk with db AND log volumes on it and the other two disks with only db volumes. Or will this layout degrade performance too much? -- Best wishes, Dirk Kastens Universitaet Osnabrueck, Rechenzentrum (Computer Center) Albrechtstr. 28, 49069 Osnabrueck, Germany Tel.: +49-541-969-2347, FAX: -2470
disk layout for db and log
Hi, we bought a new IBM @server p5 520 for TSM. The server has 8 internal 70 GB harddisks. hdisk0 and hdisk1 contain the mirrorred rootvg with the operating system. So we have six disks left for the TSM db and recovery log. Both shall be mirrored, what makes three disks. What is the best disk layout? Should we create one big db volume on each of the three disks plus one big log volume? Maybe hdisk2: db1 (65 GB) + log1 (4 GB) hdisk3: db2 (65 GB) + log2 (4 GB) hdisk4: db3 (65 GB) + log3 (4 GB) Or should the recovery log be located on a separate disk (what will be a waste of space)? Shall we create on big db volume on every disk or several small volumes? Maybe hdisk2: db1 + db2 + db3 (20 GB) + log1 (10 GB) hdisk3: db4 + db5 + db6 (20 GB) hdisk4: db7 + db8 + db9 (20 GB) Any ideas are welcome. -- Regards, Dirk Kastens Universitaet Osnabrueck, Rechenzentrum (Computer Center) Albrechtstr. 28, 49069 Osnabrueck, Germany Tel.: +49-541-969-2347, FAX: -2470
